Computational Advances in Structural Biology

Abstract:

The nuclear pore complexes (NPCs) are colossal 120-megadalton structures that mediate transport in and out of the nucleus. The NPC adopts a channel-like structure formed of a protein scaffold, which perforates the membrane, and several structurally flexible subunits that attach to the scaffold and form the selective transport barrier of the NPC. Determining this structure at the atomic level would help to understand the nucleocytoplasmic transport and many other processes such as virus invasion through the NPCs. However, due to the huge size and flexibility of the NPC, its structure could not yet be determined by a single technique such as X-ray crystallography or electron microscopy. The previous models of NPCs have thus been resolved by integrating several complementary techniques such as structural modeling, cryo-electron tomography (cryo-ET), X-ray crystallography, and crosslinking mass spectrometry. Still, the previous models remained incomplete and low precision.

Here, I will present how we built a new model of the human NPC by integrating cryo-ET, prior experimental knowledge, and AlphaFold. The resulting models have an unprecedented level of detail and completeness, comprising 70 MDa and around 90% of the structured NPC scaffold. The model describes how the NPC is organized by structured domains and flexible linkers and how it binds the membrane. The high precision has allowed us to conduct molecular dynamics simulations, and it now serves as a starting point for understanding the mechanism of transport. Our work serves as a prime example of how AI-based modeling can be combined with in situ structural biology to unravel the subcellular architecture.